Abstract
SummaryThe determination of endometrial carcinoma histological subtypes, molecular subtypes, and mutation status is critical for the diagnostic process, and directly affects patients’ prognosis and treatment. Sequencing, albeit slower and more expensive, can provide additional information on molecular subtypes and mutations that can be used to better select treatments. Here, we implement a customized multi-resolution deep convolutional neural network, Panoptes, that predicts not only the histological subtypes but also the molecular subtypes and 18 common gene mutations based on digitized H&E-stained pathological images. The model achieves high accuracy and generalizes well on independent datasets. Our results suggest that Panoptes, with further refinement, has the potential for clinical application to help pathologists determine molecular subtypes and mutations of endometrial carcinoma without sequencing.
Highlights
Endometrial cancer is the most common type of gynecologic cancer among women around the world, with an increasing occurrence and mortality.[3,4,5,6] In the United States, it is one of the top 5 leading cancer types, with 52,600 new cases reported in 2014, which increased to 61,880 in 2019.5–7 Globally, endometrial cancer caused 42,000 women’s deaths in 2005, and this annual mortality count estimate drastically increased to 76,000 in 2016.3,4 The 5-year survival rate, depending on the study cohort, ranges from 74% to 91% for patients without metastasis.[5]Clinically, endometrial carcinomas are stratified based on their grade, stage, hormone receptor expression, and histological characteristics.[8]
We show that models using this architecture can classify common endometrial carcinoma histological subtypes, molecular subtypes, and several critical mutations with decent performance based on hematoxylin and eosin (H&E) images and outcompete existing InceptionResnet models in most top-performing tasks
Cases in the mixed dataset were randomly split into training, validation, and test set, such that slides from the same patient were in only one of these sets
Summary
Endometrial carcinomas are stratified based on their grade, stage, hormone receptor expression, and histological characteristics.[8] Histological classification reflects tumor cell type and informs the choice of surgical procedure and adjuvant therapy. The majority of endometrial cancer cases exhibit either endometrioid (70%–80% of cases) or serous (10% of cases) characteristics.[9] Patients with serous subtype tumors have a lower 5-year survival rate due to more frequent metastases and a higher risk of recurrence[4] it is critical to determine the subtypes to determine patients’ individualized treatment plans and to assess prognosis.[3,10] Histological subtype is determined by pathologists after thorough examination of hematoxylin and eosin (H&E)-stained tissue sample slides of tumor samples. Endometrioid tumors typically exhibit a glandular growth pattern, while the serous subtype is characterized by the frequent presence of a complex papillary pattern.[11,12,13] These features are not exclusive for either of the subtypes, making histological classification challenging, especially among high-grade cases, even for experienced pathologists and necessitating ancillary subtyping criteria.[1,4,14,15]
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
